Hide Forgot
Description of problem: I have a RHCS two-node cluster with configuration managed/distributed by ccs/ricci. Ricci daemon is by default executed under 'ricci' user. With this settings even simple command takes tens of seconds to proceed. E.g.: $ time ccs -h localhost --lsnodes ceriha1.hb: nodeid=1 ceriha2.hb: nodeid=2 real 0m21.252s user 0m0.097s sys 0m0.028s During the execution, ricci waits most of the time for virsh to finish: ricci -u ricci ├─virsh nodeinfo └─{ricci} but virsh timeouts on communication with libvirtd through socket: ... connect(6, {sa_family=AF_FILE, path=@"/var/lib/ricci/.libvirt/libvirt-sock"}, 110) = -1 ECONNREFUSED (Connection refused) nanosleep({1, 900000000}, NULL) = 0 connect(6, {sa_family=AF_FILE, path=@"/var/lib/ricci/.libvirt/libvirt-sock"}, 110) = -1 ECONNREFUSED (Connection refused) nanosleep({2, 0}, NULL) = 0 connect(6, {sa_family=AF_FILE, path=@"/var/lib/ricci/.libvirt/libvirt-sock"}, 110) = -1 ECONNREFUSED (Connection refused) gettid() = 12788 close(6) = 0 gettid() = 12788 gettid() = 12788 gettid() = 12788 write(2, "error: ", 7) = 7 write(2, "Failed to reconnect to the hyper"..., 38) = 38 write(2, "error: ", 7) = 7 write(2, "no valid connection\n", 20) = 20 write(2, "error: ", 7) = 7 write(2, "Failed to connect socket to '@/v"..., 88) = 88 fstat(1, {st_mode=S_IFIFO|0600, st_size=0, ...}) = 0 mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fa9e7a55000 write(1, "\n", 1) = 1 exit_group(1) = ? We use default libvirt configuration and connection from common user shell works fine $ virsh nodeinfo >/dev/null; echo $? 0 If I run ricci daemon under root, i.e. change RUNASUSER="root" in /etc/init.d/ricci, the command successfully ends quickly: $ time ccs -h localhost --lsnodes ceriha1.hb: nodeid=1 ceriha2.hb: nodeid=2 real 0m0.300s user 0m0.093s sys 0m0.039s Version-Release number of selected component (if applicable): CentOS release 6.2 (Final) libvirt-0.9.4-23.el6_2.4.x86_64 ricci-0.16.2-43.el6.x86_64 ccs-0.16.2-43.el6.x86_64 - libvirt in default settings - no SElinux
This is an issue caused by changes to libvirt, we're working on determining the best place to make a fix. For now, this issue can be worked around on the cluster nodes with the following commands: mkdir /var/lib/ricci/.libvirt chown ricci.ricci /var/lib/ricci/.libvirt Thanks! Chris
Quickfix helped, thanks!
This is intended to be addressed in libvirt. See bug 785038
This fix in libvirt will not require any changes to ricci. *** This bug has been marked as a duplicate of bug 785038 ***
*** Bug 783027 has been marked as a duplicate of this bug. ***
Created attachment 572292 [details] Contains screenshots with Error and luci logs while attempting to create a new cluster.
This has been fixed by a 6.2 zstream fix in libvirt-0.9.4-23.el6_2.7. Please let me know if you see this issue after upgrading to the latest version of libvirt.